A scan is a picture of words, not words — so the text has to be recognized before it can be translated. Upload the scan and OCR extracts the text, AI translates it into any of 280+ languages, and the page layout is preserved, so what you download still looks like the document you uploaded. Phone photos of pages work too.
Used by 1,000+ enterprise brands and rated 4.6/5 on G2, for everything from legal files to technical records — the kind of content that has to be right in every language.
Translating a scanned document is a two-stage process: optical character recognition (OCR) first converts the image of the page into machine-readable text, and only then is that text translated. Smartcat runs both stages in one upload.
OCR detects the text on the scanned page, including text inside tables and headers, and reconstructs the document’s structure.
Translation then takes the extracted text into any of 280+ languages, using your glossary and translation memory so recurring names and terms stay consistent across a whole folder of scans.
The output is a document in the new language with the original layout preserved, and every segment stays editable, so an OCR misread can be corrected before it becomes a mistranslation.
Accuracy depends mostly on the scan itself: aim for at least 300 DPI, with the page straight, evenly lit and high in contrast, and clean printed text is recognized far more reliably than handwriting.
This page is for image scans: JPG, PNG, TIFF and files straight from a scanner or phone camera. Scanned PDFs, including multi-page ones, belong with the PDF translator — choose the pdfimageonly parsing method at upload, which routes each page through this same image-OCR pipeline and keeps the original layout. Every other format has its own workflow too.
Scanned PDFs use the pdfimageonly parsing method, which runs this same OCR page by page and keeps the original layout.
Ideal for quick edits to Google Docs files.
Handle data-heavy files without losing structure.
Keep presentations consistent across languages.
PPT Translator to maintain visuals and text alignment.
Translate your spreadsheets efficiently.
Translate XLIFF files for professional localization workflows.
Translate your TXT files with Smartcat easily.
Translate HTML content for your website.
A simple XML translator for structured data.
Translate the text inside images while keeping the original layout.
Translate JPG files with Smartcat's Image translation tool.
Translate PNG files with Smartcat's Image translation tool.
Translate SRT files and make subtitles for global audiences.
Upload your VTT files and translate them now.
Instantly translate your audio into multiple languages.
Translate your video files for global audiences.
Record and translate WAV files into your preferred languages.
Translate MOV files with Smartcat's MOV file translator.
Translate the text in a GIF — for an animated GIF, the first frame is translated.
for ease of setup
for ease of use
enterprise brands
average rating on G2
A 1:1 walkthrough with our team, no commitment — bring the scans you actually need translated.
300 DPI
the one number to aim for
Below that, character edges blur and OCR starts guessing. Alongside resolution, three things decide how well a page reads: the page straight rather than skewed, evenly lit with no shadow across the text, and high in contrast between ink and paper.
30 MB
per image, via OCR
A straight-on, evenly lit phone photo behaves like a scan. Skewed or shadowed shots produce errors — retaking the photo flat is faster than fixing misreads. Fax quality or a photocopy of a photocopy means manual correction in the editor, and cursive handwriting stays unreliable.
280+
target languages per upload
One scan, as many targets as you need, with one glossary across a whole folder so case numbers and names are translated identically throughout. Printed non-Latin scripts are supported — see scanned Chinese documents and scanned Arabic documents, or translate multiple images at once for volume.
Up to 70%
Cost savings
Stanley Black & Decker cut translation costs while raising quality.
31 hours
Of work time saved monthly
for Babbel’s marketing and L&D departments
3 days
to deliver what took 10
Smith+Nephew cut eLearning translation turnaround from an average of 10 days with previous providers to two to three days for the same course length.
Upload the scan, pick the language, download a document that still looks like the original. Free for 15 days with 15,000 Smartwords, no credit card, with full access to translation capabilities.
Upload the scan, set English as the target language, review the extracted text, and download the translated document. OCR converts the picture of the page into text first, then that text is translated. The whole pass takes minutes for a clean scan; budget review time for poor-quality pages or handwriting.
The scan decides most of it. Aim for at least 300 DPI, with the page straight, evenly lit and high in contrast; below that, character edges blur and OCR starts guessing. OCR errors also propagate — a misread character produces a fluent, confident, wrong translation downstream — which is why the review step exists, and why rescanning a bad page beats trusting a good-looking output from one.
The layout is preserved: tables stay tables and headers stay headers, so the translated page can be compared with the original side by side. Non-text elements such as stamps and signatures remain part of the page rather than being translated.
JPG/JPEG/JFIF, PNG, TIF/TIFF, BMP, GIF, DCX/PCX, JP2/JPC, DJVU/DJV, JB2, image-only PDF and AI files are processed via OCR, up to 30 MB per image. Single-page TIFF is explicitly supported. HEIC and WebP are not on the supported list, so convert those first. For scanned PDFs, use the PDF translator.
Yes — a straight-on, evenly lit photo behaves like a scan. Skewed or shadowed shots produce OCR errors, so retaking the photo flat is usually faster than fixing the output.
Printed text, reliably; neat block handwriting, sometimes; cursive, no — treat any handwriting output as a starting draft to verify word by word. For documents where the handwriting is the content, a human translator is the honest recommendation.
Yes, with a mechanism worth understanding. Machine output alone is not a certified translation; add a professional translator from the Marketplace in the same workflow and that translator signs a statement of accuracy covering the work they did.
Smartcat facilitates the attestation as the platform — the certification comes from the linguist, not from Smartcat, and Smartcat does not certify the translation itself or assume liability for its quality. An account manager can complete the signing manually on the linguist’s behalf where needed.
Some authorities go further and require a translator sworn or registered in their own country — the requirement behind a German beglaubigte Übersetzung, a French traduction assermentée or a USCIS filing.
Scans are usually the sensitive material — contracts, records, IDs. Smartcat is SOC 2 Type II compliant; files are encrypted in transit and at rest, workspaces are isolated, and access is role-based. Details on the security page.
Statista Market Insights. "Language translation NLP – Worldwide." Statista, March 2025.
Statista Research Department. "Number of employees in interpreting and translating services in the United States from 2011 to 2022." Statista, 2023.
Statista Research Department. "Market size of translation services in the United States from 2015 to 2022." Statista, 2023.
Liu, Z., Wu, C., & Xu, X. (2025). The Role of Artificial Intelligence in Sustainable Development and Industrial Transformation. Asia Pacific Economic and Management Review, 2(2).